Paraformer-zh-streaming

Streaming speech recognition — real-time transcription with low latency for Chinese.

Streaming variant of Paraformer that processes audio in chunks, delivering results in real-time as audio arrives.

Quick Start

from funasr import AutoModel

model = AutoModel(
    model="funasr/paraformer-zh-streaming",
    hub="hf",
    device="cuda",
)

# Streaming inference with chunk processing
import sounddevice as sd
import numpy as np

chunk_size = [0, 10, 5]  # [lookback, chunk, lookahead] in 60ms frames
result = model.generate(
    input="audio.wav",
    chunk_size=chunk_size,
    encoder_chunk_look_back=4,
    decoder_chunk_look_back=1,
)
print(result[0]["text"])

Features

• Real-time streaming recognition
• Low latency chunk-by-chunk processing
• Chinese + English support
• Compatible with WebSocket deployment

Model Details

Property	Value
Architecture	Paraformer-Streaming
Parameters	220M
Languages	Chinese, English
Mode	Streaming (online)
Sample Rate	16kHz

Links

• GitHub: FunASR
• Docs: modelscope.github.io/FunASR
• Offline version: funasr/paraformer-zh